***Setmana 4, Classe 1:

**  Repàs de regressió múltiple.  Dades de paisos.sav

*https://www.econ.upf.edu/~satorra/dades/PAISOS.SAV

*Pregunta de interés,  efecte de les calories sobre l'esperança de vida ¿

*Primer mirem la variable esperança de vida

GRAPH
  /HISTOGRAM(NORMAL)=espvida .

*observem una distribució bimodal,  es la mitjana important en aquest cas ¿

GRAPH
  /SCATTERPLOT(BIVAR)=calories WITH espvida
  /MISSING=LISTWISE .



*associació positiva 

*Cal controlar per altres variables 

*Diagrama de dispersió matricial

GRAPH
  /SCATTERPLOT(MATRIX)=espvida pib calories habmetg
  /MISSING=LISTWISE .


Gràfics com el del PIB mostren la necessitat de normalitzar 

GRAPH
  /HISTOGRAM(NORMAL)=pib .


*Transformació de variables 

COMPUTE logpib = LG10(pib) .
EXECUTE .


COMPUTE loHmetge = LG10(habmetg) .
EXECUTE .



GRAPH
  /SCATTERPLOT(MATRIX)=espvida calories logpib lohmetge
  /MISSING=LISTWISE .

Regressio simple

REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER calories
  /PARTIALPLOT ALL
  /SCATTERPLOT=(*ZRESID ,*ZPRED )
  /RESIDUALS HIST(ZRESID) NORM(ZRESID) .


*alta significació de calories


REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib
  /PARTIALPLOT ALL
  /SCATTERPLOT=(*ZRESID ,*ZPRED )
  /RESIDUALS HIST(ZRESID) NORM(ZRESID) .


*alta significació de pib

REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib calories
  /PARTIALPLOT ALL
  /SCATTERPLOT=(*ZRESID ,*ZPRED )
  /RESIDUALS HIST(ZRESID) NORM(ZRESID) .


*significació de calories 
 
REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib calories lohmetge
  /PARTIALPLOT ALL

*no significació de la variable calories


*Posem variables dummies de nivell de desenvolupament 

COMPUTE nivell1 =  nivell = 1 .
EXECUTE .


COMPUTE nivell2 =  nivell = 2 .
EXECUTE .

COMPUTE niv1cal = nivell1 * calories.
EXECUTE .


COMPUTE niv3cal = nivell3 * calories.
EXECUTE .


EXECUTE .
COMPUTE nivell3 = nivell = 3
 .
EXECUTE .
COMPUTE niv3cal =  nivell3 * calories
 .
EXECUTE .


REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib calories lohmetge   nivell3 niv3cal
  /PARTIALPLOT ALL
  /SCATTERPLOT=(*ZRESID ,*ZPRED )
  /RESIDUALS HIST(ZRESID) NORM(ZRESID) .




*Finalment tenim una resposta sobre l'efecte de les calories que és matisada. 


*** En general,  un exemple d'anàlisi de la covariancia es 

REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA CHANGE
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib calories lohmetge  /METHOD=ENTER nivell1 nivell2
  nivell3  .


*** Dummy variables    (diferencia en espvida en els diferents continents)
COMPUTE Africa = cont=2 .
EXECUTE .

COMPUTE Asia= cont=3 .
EXECUTE .

COMPUTE America = cont=4 .
EXECUTE .

COMPUTE Oceania = cont=5 .
EXECUTE .

******* ANOVA 

REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA CHANGE
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
 /METHOD=ENTER africa asia america oceania  .

******** ANCOVA  (regressio per blocks ) 

REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA CHANGE
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib  /METHOD=ENTER africa asia america oceania  .



COMPUTE Yeur =   (37.8 + 0)+   9.17*logpib  .
EXECUTE .

COMPUTE Yafric =  ( 37.8 + -8.877) +   9.17*logpib  .
EXECUTE .

COMPUTE Yasia =  ( 37.8 - .79)  +   9.17*logpib  .
EXECUTE .

COMPUTE Yame =   (37.8  + 1.7) +   9.17*logpib  .
EXECUTE .

COMPUTE Yame =   (37.8  + 1.7) +   9.17*logpib  .
EXECUTE .

COMPUTE Yoce =   (37.8  + .55) +   9.17*logpib  .
EXECUTE .





*** representacio grafica ancova

GRAPH
  /SCATTERPLOT(OVERLAY)=     logpib logpib logpib      logpib   WITH yeur yafric yasia yame  
(PAIR)
  /MISSING=LISTWISE .




********* ANCOVA amb efectes de interaccio

COMPUTE AfrPIB=  logpib*africa .
EXECUTE .

COMPUTE amerPIB=  logpib*america.
EXECUTE .


REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA CHANGE
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib  /METHOD=ENTER africa asia america oceania  /METHOD=ENTER  amerPIB AfrPIB  .


COMPUTE Yeur =   (41.534  + 0)+   8.255*logpib  .
EXECUTE .

COMPUTE Yafric =  ( 41.534   -23.539) +   (8.255 + 4.99  )*logpib  .
EXECUTE .

COMPUTE Yasia =  ( 41.534   - 1.3)  +   (8.255  )*logpib  .
EXECUTE .

COMPUTE Yame =   (41.534  + 5.98) +   (8.255 - 1.46 )*logpib  .
EXECUTE .

COMPUTE Yoce =   (41.534   -.12) +   (8.255  )*logpib  .
EXECUTE .

 

*** representacio grafica ancova amb interaccio

GRAPH
  /SCATTERPLOT(OVERLAY)=     logpib logpib logpib      logpib   logpib WITH yeur yafric yasia yame yoce 
(PAIR)
  /MISSING=LISTWISE .


REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT espvida
  /METHOD=ENTER logpib  /METHOD=ENTER asia america oceania africa
 /METHOD=ENTER afrpib amerpib
  /SAVE COOK LEVER RESID SRESID .

COMPUTE e2 =   res_1*res_1  .
EXECUTE .



GRAPH
  /SCATTERPLOT(BIVAR)=coo_1 WITH e2
  /MISSING=LISTWISE .


GRAPH
  /SCATTERPLOT(BIVAR)=coo_1 WITH sre_1
  /MISSING=LISTWISE .



GRAPH
  /SCATTERPLOT(BIVAR)=coo_1 WITH sre_1
  /MISSING=LISTWISE .




GRAPH
  /SCATTERPLOT(BIVAR)=coo_1 WITH sre_1
  /MISSING=LISTWISE .